Say more about "couldn't even turn [it] on"? Because it didn't work then (not a good sign)? Or because they, or somebody, thought the vortex would detach and go rogue?
It was less of an immediate safety concern and more a case of trying to limit the variables. They were only doing tests under ideal conditions, rather than testing in all conditions. I imagine the plan involves testing in all conditions eventually, but they need to fine-tune their equipment in ideal conditions to make sure they know what "good" and "bad" look like before potentially feeding some atmospheric action that they didn't anticipate.
How about doing it somewhere remote? Say, on a remote pacific island with no inhabitants? Or is there even the fear that it could emit a hurricane (which would probably happen anyways, but if it's your machine that starts it, all (media) hell breaks loose)?